Arcee Small Language Models

Arcee AI started as a model training company, assembling a world class team of researchers who pioneered model post-training techniques and open source libraries such as Spectrum for efficient model training, MergeKit for Model Merging, and DistillKit for Model Distillation. We have taken these research advancements and applied the techniques to leading open source models to create state-of-the-art small language models (SLMs).

Arcee defines SLMs as a model which can be run efficiently on a single GPU instance, allowing for deployment of the models in your own environment. Our models range from 150M to 72B parameters and are optimized for cost efficiency and performance while maintaining high levels of accuracy.

Arcee SLMs are fine-tuned to be task specific, offering general purpose, reasoning, coding, function calling, and vision models. Below is a high level overview of our models:

General Purpose

• Arcee Blitz

• Virtuoso Small

• Virtuoso Medium

• Virtuoso Large

Reasoning

• Maestro

Coding

• Coder Large

Function Calling

• Caller Large

See Model Selection for more detailed information on each model.